✕

• Landing page with hero, features, pricing, and signup/login
• User accounts with session tokens and 7-day expiry
• Dashboard with credits, analysis history, account settings
• Checkout page with plan selection and auth gate
• Premium add-ons section (coming soon)
• Industry-agnostic term matching — removed all hardcoded lists, AI handles all classification
• Validation logic moved server-side (api.php) to protect IP
• .htaccess blocks direct access to JSON/log/env files
• Job URL fetch with Totaljobs support, Indeed/LinkedIn graceful blocking
• All logos now link back to landing page

• Job URL import now targets main job-description content more tightly instead of broad page text
• Fetch button styling softened so it is less visually dominant than the main CTA

• Step 2 URL tab now fetches job descriptions directly from job listing pages
• Added one-click import into the JD textarea from supported URLs
• Graceful error handling for invalid URLs, blocked pages, and unusable extracts

• Added Single and Bundle pricing cards to the unlock screen
• Bundle card now uses highlighted yellow badge and premium border treatment
• Single card now has a grey Standard badge for clearer plan hierarchy
• Screen 4 and 5 restored inside the main content flow so they inherit normal page padding
• Pricing cards now align to the same height with CTA buttons aligned horizontally
• Mobile layout improved for pricing cards, screen padding, and bottom navigation anchoring
• CTA buttons now respect card padding correctly

• Complete architecture overhaul — structured parse → classify → rewrite → validate → compare
• Match classification: Evidenced / Inferable / Unsupported — stops fabricating unsupported skills
• Multi-factor scoring across 6 dimensions: Relevance, Evidence, Quality, ATS Parseability, Credibility, Improvement
• Confidence badge — shows if AI is confident the rewrite is genuinely stronger
• Score breakdown bars — see exactly where the CV is strong and weak
• Gap reporting — unsupported requirements surfaced honestly instead of being forced into CV
• Output validation — rejects placeholders, duplication, broken formatting before display
• 5-step AI workflow: Parse → Classify → Score → Rewrite with constraints → Self-check
• "Do not invent" rule — AI cannot add tools, methods, or responsibilities not in the CV
• Section-by-section improvement instead of appending new content on top
• Killed n-gram extraction — no more garbage phrase matching

• Upgraded to GPT-4o (from mini) for significantly better analysis quality
• Defined senior recruiter persona for more expert-level feedback
• Added explicit scoring rubric (0-100) for consistent scoring
• AI now verifies local analysis hints instead of trusting blindly
• Enforced accuracy — only claims matched if verifiably in CV text
• Checks synonyms, abbreviations, slash variants before marking missing
• Max limits on all output arrays for focused, relevant feedback

• Removed n-gram extraction — no more garbage phrases in matched/missing chips
• Removed partial matching — only exact and synonym matches count
• Matched strengths now only show terms verifiably present in your CV
• Missing from CV now only shows terms genuinely absent
• Slash/compound variant matching (e.g. UX Design matches UX/UI Design)

• ATS readiness legend — colour-coded score bands (Red/Amber/Blue/Green)
• Score bars now dynamically coloured by ATS readiness threshold
• AI quality rules enforced: no duplication, no generic filler, enhance over stack
• Client-side dedup filter — removes generated content overlapping with original CV
• Generic phrase blocklist — filters out weak professional filler
• Removed exposed API key from client-side code
• Removed debug console logging
• Fixed blank unlock page (screen 4) — missing </div> in peek card

• Analysis now powered by OpenAI — scoring, summary, skills, bullets all AI-generated
• Local API server runs GPT-4o-mini for fast, quality CV tailoring
• Loading animation with progress steps while AI analyses
• AI prompt enforces all quality rules — no broken fragments, no keyword stuffing
• Preserves candidate's original voice — only tailors when confident
• AI-generated scores for current match and optimised potential
• Fallback to local analysis if API server is unavailable
• Extraction quality gate prevents broken fragments from entering output
• Skills validated as standalone professional labels before display

• Extraction quality gate — rejects broken fragments, weak words, repeated stems
• Summary preserves candidate's original voice — only lightly tailors
• Skills validated as standalone professional labels before output
• Experience bullets kept intact — no more awkward term appending
• Download uses same clean pipeline as preview

• Rewrites now produce clean, human-readable prose — no more broken keyword fragments or stitched phrases
• Professional summary uses natural sentence structure a recruiter would actually read
• Core skills are curated, deduplicated, and comma-separated — no raw term dumping
• Experience bullets no longer awkwardly append missing terms
• Download output matches the improved quality of the preview

• Rewrite output no longer produces broken keyword fragments or stitched phrases
• Summary and skills sections now read like natural recruiter-facing prose
• Core skills are curated and clean — no awkward keyword stacking
• Critical issues, quick wins, and rewrite highlights sections now populate reliably on analysis

• Added dedicated critical issues section for genuine ATS blockers
• Added quick wins section for fast visible score improvements
• Added rewrite highlights section with stronger tailored wording suggestions
• Improved results layout so issues, wins, and rewrites are easier to scan
• Sharper distinction between missing terms and high-priority ATS gaps

• Term categories: hard skills, tools, certifications, soft skills, qualifications — scored independently
• ATS-weighted scoring: tools & hard skills 4–5×, soft skills only 1.5×
• Potential score capped at 92 — over-optimisation is a real ATS penalty
• Matched/missing chips now show only ATS-relevant terms — no random fragments
• N-gram extraction limited to requirement lines only — eliminates noise
• Removed all guaranteed/outcome language — scores are estimates, not promises
• Deterministic potential calculation — no randomness

• Deterministic scoring, trigram extraction, position-weighted analysis
• Rewritten experience bullets, smarter summary generation

• Dynamic ATS scoring, synonym-aware matching, CV peek, download, 4-step wizard